Eyes on the Road: A Machine Learning-based Fatigue Detection System for Safer Driving

Abstract

A driver's drowsiness can cause accidents, but a car safety technology called a fatigue detection system can stop them. One of the main causes of accidents on the road is drowsy or sleepy driving. There is an increase in the number of fatalities and other serious injuries every year worldwide. This project intends to improve driver drowsiness detection implementation and further optimise driver drowsiness detection by developing a fatigue detection system using machine learning that takes driver drowsiness data and additional viewpoints on the aforementioned issues into account. In this study, the supervised machine learning algorithm PERCLOS, which is taught using a dataset by identifying eye movements, is employed. The developed system first captures the image of a user using a camera. Next, it uses various python modules like NumPy and OpenCV to detect the eyes of a user. After detecting the eyes of a user, it feeds this image to obtain the ratio of the eye shape and the distance between the top and bottom endpoints. The system then considers the aspect ratio of eyes and uses default base case to be 25%. The system also detects the moments of a user, and if the there is no moment found by the system within next four seconds, it sends an alert in the form of an alarm beep, which a user can stop and then the system will be directed to the start of the cycle.